Loudness presentation by Orban

Optimum TechnologiesOptimum Technologies

Automatic Loudness Control in Television Broadcast:

Orban’s Implementation ofthe CBS Loudness Meterand Loudness Controller


Automatic Loudness Control Has a Long History

• The first automatic loudness control technology was developed by CBS Laboratories in the mid-1960s in response to a Federal Communications Commission study regarding audience complaints about objectionably loud commercials.

• Jones and Torick at CBS Technology Center revisited this work in 1981 to improve loudness meter accuracy. This work was published in the SMPTE Journal.


Automatic loudness control starts with a loudness meter

whose indications match subjective loudness as perceived by listeners.


Loudness Meter must take into account:

• Loudness Addition: For a given total sound power, the sound becomes louder as the power is spread over a larger number of critical bands (about 1/3-octave).

• Frequency Dependence: The ear’s perception of loudness is strongly dependent on frequency.

• Loudness Integration in Time: A given amount of acoustic power sounds progressively louder until its duration exceeds about 200 milliseconds, at which point no further loudness increase is heard.


Loudness Meter Technology

• Loudness Addition: The meter first divides the signal into frequency bands and applies each band to a rectifier followed by a fast averaging.

• Frequency Dependence: The averaged outputs of the bands are summed with unequal gains that mimic the frequency-dependence of the ear.

• Loudness Integration in Time: The sum of the smoothed filter outputs is applied to a filter with an integration time of approximately 200 ms.


Loudness Meter Block Diagram


Loudness Meter Filterbank


Filterbank Curves & Summation


Loudness Meter Accuracy Limitations 1

• Loudness meter accuracy is inherently limited by the fact that human listeners disagree by as much as 4 dB when asked to match the loudness of test program material with a reference tone or wideband noise. Different people perceive loudness differently.

• A loudness meter can only be calibrated for a fixed acoustic listening level because the ear’s sensitivity as a function of frequency is level-dependent.


Loudness Meter Accuracy Limitations 2

• The room acoustics and frequency response of the listener’s playback system are unpredictable.

• These issues mean that automatic loudness measurement and control for broadcast will always be approximate.


Comparison ofLeq(RLB) and CBS Algorithms• Leq(RLB) takes into account frequency

dependence but not loudness addition. It does not consider the loudness integration time constant of the ear, so it can only measure “long-term” loudness by creating a single loudness measurement for program segments exceeding ~3 seconds in length.

• The CBS algorithm, by taking into account all three factors, is useful with both speech and wideband non-speech material and can indicate short-term, potentially annoying loudness peaks such as “esses” in speech.


Automatic Loudness Control

• To make an automatic loudness controller, one can insert an loudness meter into a servo loop where the loudness meter monitors the loudness controller’s output and constrains it to a preset threshold by applying gain reduction as needed.

• The servo can be designed to produce wideband or multiband gain reduction before the loudness meter. Appropriate multiband design minimizes audible gain pumping.


Loudness Control in Practice

• CBS’s 1981 loudness controller technology was licensed to Orban and CRL.

• There are thousands of processors using this technology on-air in television stations (mostly analog) throughout the world.

• Experience has shown that using this technology significantly reduces listener complaints caused by loud commercials.


Automatic Loudness ControllerProgram Context Limitations

• An automatic loudness controller operates with reference to an absolute subjective loudness threshold that does not adapt to program context as well as a human mixer.

• For example, if there is a transition between very quiet program material (like footfalls through rustling leaves or quiet underscoring) and a commercial, the commercial may still seem offensively loud even though the Loudness Controller is controlling its loudness correctly with reference to other sounds that reach full-scale loudness.

• While automatic speech/non-speech discrimination can help a loudness controller understand context, it cannot deal with all situations (like the examples above, where adjacent elements are both “non-speech”).


Loudness Measurements for This Presentation

• The 1981 CBS Loudness Indicator was used to make the following measurements.

• The audio processor was an Orban Optimod-DTV 8585.


Loudness Measurements

• The measurements are divided into 5-second bins that contain the highest meter indication in a given 5-second interval.


Loudness Measurements

• Program material is a 14-minute recording from the output of a network-affiliated TV station’s master control switcher.

• The material consists of a daytime drama interspersed with commercialsand programming promotions.


Loudness Measurements• The following measurements show that the

loudness consistency of the unprocessed feed is not satisfactory.

• Listening tests verify this. Program material loudness is annoyingly inconsistent and commercials are much louder than the program material. In the following chart, the commercial clusters can be easily recognized just because of their higher and more consistent loudness.


Unprocessed Feed

Time (seconds)0 200 400 600 800

Rel

ativ

e lo

udne

ss in

dB

(CB

S A

lgor

ithm

, 198

1)

-20.0

-15.0

-10.0

-5.0

0.0

5.0

Loudness vs. Timeof Unprocessed Audio from Master Control Switcher Output


Unprocessed Feed

Relative Loudness in dB (CBS algorithm, 1981)Each bin is maximum loudness in 5-second interval

-25.0 -20.0 -15.0 -10.0 -5.0 0.0 5.0

Num

ber o

f bin

s (o

ut o

f 168

tota

l bin

s)

0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

40.0

Histogram:Unprocessed Audio from Master Control switcher


Audio Processing for Loudness Control

• The next measurements shows the loudness consistency created by 2-band compression without loudness control. The audio processor was running its TV 2-Band preset.

• Consistency, while improved, is still not good enough to prevent viewer annoyance.

• Note that the loudness scale in the graphs has changed compared to the unprocessed measurement.


2-Band Compression: no LC

Time (seconds)0 200 400 600 800

Rel

ativ

e Lo

udne

ss in

dB

(CB

S A

lgor

ithm

; 198

1)

-10.0

-5.0

0.0

5.0

Loudness Control using 2-Band Compressor (each point is maximum loudness in 5-second bin)Souce Audio is material is drama interspersed with commercials.


2-Band Compression: no LC


-10.0 -5.0 0.0 5.0

Num

ber o

f bin

s (o

ut 1

68 to

tal b

ins)

0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

40.0

Histogram:2-Band Compression without Loudness Controller


Audio Processingfor Loudness Control

• The next measurements show the loudness consistency created by 2-band compression with automatic loudness control(TV 2B+LC preset).

• Automatic loudness control significantly improves results, preventing “esses” and commercials from sounding objectionably loud.


2-Band Compression + LC

Time (seconds)0 200 400 600 800

Rel

ativ

e Lo

udne

ss in

dB

(CB

S A

lgor

ithm

; 198

1)

-10.0

-5.0

0.0

5.0

Loudness Control 2-Band with CBS Loudness Controller; (each point is maximum loudness in 5-second bin)Souce Audio is material is drama interspersed with commercials.



Relative loudness in dB (CBS algorithm, 1981)Each bin is maximum loudness in 5-second interval

-10.0 -5.0 0.0 5.0

Num

ber o

f bin

s (o

ut o

f 168

tota

l bin

s)

0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

40.0

Histogram:2-Band Compression with CBS Loudness Controller



Time (seconds)0 200 400 600 800R

elat

ive

Loud

ness

in d

B (C

BS

Alg

orith

m, 1

981)

-20.0

-15.0

-10.0

-5.0

0.0

5.0

Comparison of Unprocessed Audiowith Audio Processed by 2-Band Compressor with CBS Loudness Controller

LegendUnprocessed AudioTwo Band with Loudness Controller


AGC + 5-Band Compression• Good loudness control is also possible with 5-

band compression preceded by a 2-band AGC.

• Not just any multiband compressor will work. Every aspect of the compressor - thresholds, control loop dynamics, and crossover design –must be carefully tuned to achieve effective loudness control.

• The audio processor’s TV 5B General preset was used for the measurements.


AGC+5-Band Compression

• Excessive loudness is often caused by large amounts of midrange energy that isadded in production mixing to try to make the audio “pop.” Commercials are notorious in this regard.

• Program material that has not been de-essedcan also cause objectionable loudness peaks.


AGC+5-Band Compression• 5-Band compression automatically re-equalizes

the spectral balance to make the program more comfortable for the audience to hear – it tames harsh sibilance and dental-drill midrange.

• 5-Band compression is less likely than 2-Band compression to introduce pumping caused by spectral gain intermodulation.

• The slow AGC controls average levels applied to the 5-band compressor, preventing the compressor from unnaturally increasing audio density.


AGC + 5-Band Compression

Time (seconds)0 200 400 600 800

Rel

ativ

e Lo

udne

ss in

dB

(CB

S A

lgor

ithm

; 198

1)

-10.0

-5.0

0.0

5.0

Loudness Control using 2-Band AGC Feeding Fast, Light 5-Band Compression (each point is maximum loudness in 5-second bin)Souce Audio is material is drama interspersed with commercials.


AGC + 5-Band Compression


-10.0 -5.0 0.0 5.0

Num

ber o

f bin

s (o

ut o

f 168

tota

l bin

s)

0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

40.0

Histogram:2-Band AGC Driving Fast, Light 5-Band Compressor


Comparing Processing• 2-Band compression: Does not control loudness

well enough to avoid viewer annoyance in TV audio.• 2-Band compression + CBS Loudness Control:

Effective loudness control that does not re-equalize midrange frequencies.

• AGC+5-Band compression: Effective loudness control that also prevents audible gain pumping caused by spectral gain intermodulation.

• AGC+5-Band compression + CBS Loudness Control: Most consistent loudness control; also prevents audible gain pumping caused by spectral gain intermodulation.


Processing Comparison

Time (seconds)0 200 400 600 800

Rel

ativ

e Lo

udne

ss in

dB

(CB

S A

lgor

ithm

; 198

1)

-10.0

-5.0

0.0

5.0

Loudness Control using Three Processing Styles: 2-Band, 2-Band with CBS Loudness Controller; 5-Band: (each point is maximum loudness in 5-second bin)Souce Audio is material is drama interspersed with commercials.

LegendTwo-Band, no Loudness ControllerTwo Band with Loudness ControllerFive Band


Conclusions 1:

• CBS/Orban automatic loudness control technologyeffectively controls loudness but limits dynamic range as an inevitable side-effect. Because of careful choices of compression ratios, time constants, and other subtle aspects of the processing algorithm, Orban processing limits dynamic range in a graceful way that is not noticed by listeners, who subliminally enjoy the resulting smooth, easy-to-listen-to texture.


Conclusions 2:

• In digital television, some program material is well mixed and does not require automatic loudness control. Automatic loudness control can be bypassed for such material.

• Other material, like live news and sports, requires automatic loudness control to prevent viewer annoyance – time pressures in live broadcasts prevent the audio from being carefully produced.

• Automatic loudness control is unlikely to ever be as good as a human mixer when the most esthetically pleasing results are desired. Only humans can understand the subtleties of context. But CBS/Orban automatic loudness control can please listeners and prevent annoyance-driven tune-outs caused by inconsistent loudness.

Loudness presentation by Orban

Technology

Transcript of Loudness presentation by Orban